The synapsin domain E accelerates the exoendocytotic cycle of synaptic vesicles in cerebellar Purkinje cells

Information transfer between neurons is controlled by an efficient process of regulated exocytosis of neurotransmitters stored in synaptic vesicles (SVs). In nerve terminals, SVs are organized in distinct functional pools, a large reserve pool (RP) in which SVs are restrained by the actin-based cytoskeleton, an actively recycling pool and a quantitatively smaller readily releasable pool (RRP) in which SVs are free to approach the presynaptic active zone and fuse with it upon stimulation (Rizzoli and Betz, 2005). Synapsins, a family of abundant SV-associated phosphoproteins (Greengard et al., 1993; Hilfiker et al., 1999), play a prominent role in the regulation of SV trafficking and their involvement in human pathology, namely epilepsy, has been recently recognized (Garcia et al., 2004). Synapsins regulate the pre-docking steps of exocytosis by anchoring SVs to the actin cytoskeleton and regulating their availability for exocytosis through their phosphorylation-induced dissociation from SVs and/or actin filaments (Benfenati et al., 1992a; Benfenati et al., 1992b; Ceccaldi et al., 1995; Jovanovic et al., 1996; Hosaka et al., 1999; Chi et al., 2001; Chi et al., 2003; Bonanomi et al., 2005). Moreover, synapsins have been reported to play additional roles in the later steps of neurotransmitter release, including priming and fusion (Hilfiker et al., 1998; Hilfiker et al., 2005; Humeau et al., 2001).

Synapsins are encoded by three genes (SYN1, SYN2 and SYN3) in mammals (Südhof et al., 1989; Porton et al., 1999) and alternative splicing of these genes gives rise to at least eight different isoforms composed of a mosaic of individual and common domains, the latter of which (domains A, C and E) are highly conserved evolutionarily from invertebrates to man (Kao et al., 1999). In particular, the C-terminal domain E, shared by all A-type synapsin isoforms, plays a fundamental role in synapsin function. Indeed, it is implicated in both the pre-docking and the post-docking steps of synaptic vesicle exocytosis (Pieribone et al., 1995; Hilfiker et al., 1998; Hilfiker et al., 2005) and appears to be required for the correct targeting of the synapsins to synaptic vesicles (Gitler et al., 2004a). However, thus far the exact role of domain E in neuroexocytosis has never been investigated in mammalian synapses. In synapsin knockout mice epilepsy seems to be caused by a specific deficit in inhibitory transmission (Li et al., 1995; Rosahl et al., 1995; Terada et al., 1998; Gitler et al., 2004b) and domain E is deleted in the truncation mutant of synapsin I found in X-linked human epilepsy (Garcia et al., 2004), it therefore seemed important to investigate the precise role of the synapsin domain E in mammalian inhibitory synapses.

To this aim, we generated cell-specific transgenic mice that selectively overexpress the most conserved part of the E domain (pepE) in cerebellar Purkinje cells (PCs). The results indicate that, notwithstanding a decrease in the RP of SVs, the endogenous expression of this peptide increases the performance of the inhibitory PC deep cerebellar nuclei (DCN) synapse by accelerating the kinetics of neurotransmitter release and making the synapse less susceptible to depression.

The C-terminal part of the synapsin domain E is remarkably conserved among mammalian A-type isoforms and in synapsin orthologues throughout the animal kingdom. To address the question of the functional role of this sequence in mammalian synapses, we specifically overexpressed it in a single neuronal population. A minigene was made by hybridization of two complementary oligonucleotides and cloned into the PC-specific expression vector L7ΔAUG. When expressed, the minigene was expected to induce a 25-mer peptide, corresponding to the last 25 residues of domain E from rat and mouse synapsins Ia (Fig. 1A). After injection of the L7-pepE construct into zygotes derived from the hybrid mouse strain C57Bl6×DBA2, four positive transgenic mouse lines (Tg-1 to Tg-4) were identified by PCR analysis (Fig. 1B). Two lines (Tg-1 and Tg-2) were amplified, crossed to homozygosity and bred through at least eight generations. Reverse transcriptase (RT)-PCR analysis on total RNA extracted from cerebellum, cerebral cortex and corpus striatum demonstrated that the L7-pepE transgene was expressed only in the cerebellum of both lines of L7-pepE transgenic mice (Fig. 1C and data not shown). Consistently, in situ hybridization experiments revealed that the transgene mRNA was specifically, abundantly and homogeneously expressed in the PC layer of the cerebellar cortex (Fig. 1D). No detectable expression was observed in any other brain region, except for the retina (not shown), and an identical pattern of expression was observed in both Tg-1 and Tg-2 mice.

The gross morphology of the brains from transgenic mice was virtually identical to that of their control littermates. Basic histological analysis with Nissl staining showed that the cerebellum of transgenic mice was normal in size and external appearance with regular foliation and layering of the cerebellar cortex (Fig. 2A, panels a,b). As in control animals, PCs of L7-pepE mice specifically labelled for calbindin D-28K had a characteristic morphology and displayed a typical dendritic tree in the molecular layer, the expected cell-body density in the PC layer and regular projections to the DCN (Fig. 2A, panels c-f). Thus, pepE overexpression in PCs had no detectable effect of cerebellar structure and PC differentiation and maturation.

When cerebellar sections were double stained with calbindin D-28K and anti-peptide E or anti-synapsin I antibodies, synapsin immunoreactivity exhibited a comparable distribution pattern in both transgenic and control mice, with intense labelling of the molecular and granule cell layer, `negative' appearance of the PC somata (Fig. 2B, panels a-d) and intense punctate staining in the DCN (Fig. 2Be-h; Fig. 2C). Moreover, the number of synaptophysin- and calbindin-positive PC terminals impinging on DCN neurons was virtually unchanged in L7-pepE mice (Fig. 2C,D). Although it was not possible to distinguish between pepE and endogenous synapsin, no pepE immunoreactivity was detected in PC somata, suggesting that the recombinant peptide was transported to nerve terminals together with endogenous synapsins (Gitler et al., 2004a).

To demonstrate unambiguously that pepE is expressed and correctly targeted to axonal and nerve terminal domains in transgenic PC, we enriched PCs from the cerebellum of embryonic day (E)18 Tg-2 mice and analyzed the expression of synapsin I and pepE by immunocytochemistry and immunoblotting. Immunoblotting confirmed that both synapsin I and pepE are expressed in PC-enriched cerebellar cultures; in addition, the absence of both proteins in PC cell bodies indicates that pepE is transported to nerve terminals (Fig. 3A,B). The demonstration that this is indeed the case was obtained by transfecting primary hippocampal neurons with either GFP or the GFP chimeras of synapsin Ia or domain E (Fig. 3C,D). Synapsin Ia or domain E GFP chimeras revealed a punctate distribution pattern largely overlapping with the immunoreactivity for the synaptic vesicle protein VAMP2, whereas the distribution of GFP alone was fully cytosolic.

Mice in both lines were viable, fertile and showed weight, growth, life expectancy and general motor activity comparable with wild-type littermates. Visual inspection of Tg-1 and Tg-2 mice did not reveal any gross alteration in behaviour with respect to wild-type mice. Footprints of L7-pepE mice showed no evidence of an ataxic phenotype. Step width, step length and maximum difference in stride length in transgenic mice were not statistically different from those of control littermates (Fig. 4A).

In the horizontal bar test, which evaluates dynamic balance, both wild-type and transgenic mice performed well, whereas in the vertical bar test, which evaluates static balance, both Tg-1 and Tg-2 mice obtained a significantly lower score than wild-type mice (Fig. 4B). Although transgenic animals did not fall during the 60 seconds of the test, they stayed gripped to the pole, whereas wild-type animals climbed up and down and finally got off the pole.

The ability to maintain balance on a rotating cylinder and to adapt to the rate of locomotor activity was assayed by using the rotarod test. In the constant speed mode, wild-type and transgenic mice performed equally well when the rotation speed was slow or moderately high (up to 30 rpm). However, at the maximum speed (40 rpm) both Tg-1 and Tg-2 mice displayed a statistically significant impairment in motor performance (Fig. 4C). Although wild-type mice exhibited an increased performance along successive training sessions in the accelerating mode, Tg-2 mice displayed no significant improvement in motor performance (Fig. 4C, inset). These results indicate that pepE overexpression resulted in a mild, but significant, motor impairment in balance, coordination and motor learning that becomes apparent under the most demanding conditions.

To determine whether the impairment in motor performance observed in L7-pepE mice was generated by a modification at the level of PC-DCN synapse, we analyzed neurotransmitter release from PC.

Release experiments were conducted on DCN synaptosomes labelled with [³H]GABA and stimulated with KCl, known to evoke voltage-gated Ca²⁺-channel-dependent release, mainly involving the RRP of SVs, or with ionomycin, known to diffusely increase intrasynaptosomal Ca²⁺ concentration and induce massive SV exocytosis, involving the RP of SVs (Ashton and Dolly, 2000; Stigliani et al., 2003). As shown in Fig. 5A, KCl-induced [³H]GABA overflow was similar in synaptosomes from wild-type, Tg-1 or Tg-2 mice. By contrast, ionomycin-induced [³H]-GABA overflow was significantly reduced by about 30-40% in both lines of transgenic mice with respect to control mice (Fig. 5B). These results indicate that overexpression of pepE reduces the availability of reserve SVs in PC nerve terminals, whereas the RRP of SVs seems not to be impaired.

Since PC neurons are known to experience sustained high-frequency firing, a condition that could involve slow SV cycling between the RP and the RRP, synaptosomes were stimulated with three sequential depolarizing pulses with 15 or 35 mM KCl. As expected, in wild-type synaptosomes, [³H]GABA overflows decreased in response to the progressive stimuli, as a result of SV depletion and decreased [³H]-GABA specific activity of neurotransmitter quanta. Surprisingly, GABA overflow in Tg-1 and Tg-2 synaptosomes was significantly less depressed after repetitive pulses (Fig. 5C), indicating that pepE-overexpressing PC terminals have a more efficient refilling of the RRP following intense repetitive stimuli.

To investigate synaptic transmission at the PC-DCN synapse, we performed electrophysiological experiments in acute cerebellar slices using whole-cell patch-clamp recordings from postnatal day (P)8-P12 wild-type and Tg-2 mice. DCN neurons showed a continuous barrage of spontaneous inhibitory postsynaptic currents (IPSCs) (wild-type mice, 33.6±12.9 IPSCs/second; n=9) (Tg-2 mice, 50.8±12.8 IPSCs/second; n=7; P=0.25 compared with level in wild type) reflecting tonic discharge in PCs (Anchisi et al., 2001). Since evoked IPSCs (wild type, 754.9±303.5 pA; n=9) (Tg-2, 679±241.7 pA; n=7; P=0.85) were larger than spontaneous IPSCs (wild type, 94.0±25.8 pA; n=9) (Tg-2, 51.8±9.0 pA; n=7; P=0.2), they were multi-fiber in nature (Telgkamp and Raman, 2002).

In a first set of experiments, PC axons were stimulated with a train of 20 impulses at 100 Hz to assess processes involving fast SV cycling (Fig. 6A). In control slices, IPSC amplitude declined with an exponential time course (Fig. 6B), revealing short-term synaptic depression very similar to that reported previously (Telgkamp and Raman, 2002; Pedroarena and Schwarz, 2003; Telgkamp et al., 2004). The same stimulation paradigm applied on Tg-2 slices elicited smaller depression (Fig. 6A), with a slower rate and a higher steady-state level with respect to wild-type slices (Fig. 6B). As shown in Fig. 6C, the time constant and the steady-state level of depression increased threefold and twofold, respectively, in Tg-2 versus wild-type synapses. The depression time constant changed from 13.9±2.3 milliseconds (n=4) in the wild type to 45.7±8.2 milliseconds (n=4) in Tg-2 mice (P<0.05, unpaired t-test). The steady-state level changed from 12.3±5.0% (n=4) in the wild type to 22.9±1.5% (n=4) in Tg-2 mice (P<0.05, unpaired t-test). The presynaptic nature of depression was supported by the appearance of failures in the last IPSC in the trains (wild type, 56.6±10.4%; n=4) (Tg-2, 23.1±12.9; n=3; see also below), as previously reported (Telgkamp and Raman, 2002; Pedroarena and Schwarz, 2003; Telgkamp et al., 2004).

In a second set of experiments, PC axons were stimulated at 50 Hz for 120 seconds. During such prolonged trains, which should involve slower and longer pathways of SV recycling, depression was also slower and less intense in Tg-2 compared with wild-type synapses (Fig. 7A,B). The steady state of mean IPSC amplitude was 1.95±0.4% (n=5) in wild-type synapses and 3.96±1.8% (n=5) in Tg-2 synapses (P<10^–11, ANOVA). Depression was associated with a marked increase in IPSC failures, which caused a lower steady state than in short trains (cf. Fig. 5B) supporting the presynaptic nature of the difference (Fig. 7B). The steady state of failure frequency was 83.4±3.8% (n=5) in wild-type and 69.0±3.9% (n=5) in Tg-2 synapses (P<10^–7 using χ² test).

It should be noted that in these experiments DCN neurons were maintained at negative membrane potential. Under these conditions, the previously reported forms of long-term synaptic plasticity of the PC-DCN synapse (Morishita and Sastry, 1996; Ouardouz and Sastry, 2000; Aizenmann et al., 1998) were very unlikely to occur because they require membrane depolarization. Moreover, to rule out the possibility that pepE overexpression affects PC spike firing, PC recordings performed at P15 revealed comparable spontaneous PC discharge frequencies in wild-type (5.05±0.67 Hz; n=12) and Tg-2 L7-pepE (5.22±0.40 Hz; n=14) slices (P=0.83, unpaired Student's t-test).

A further indication for a role of pepE in the regulation of RRP dynamics came from measurements of the time course of single IPSCs and of the synaptic delay in PC-DCN synapses of wild-type and Tg-2 mice. As shown in Fig. 8A,C, the IPSC rise and decay times were significantly reduced in transgenic nerve terminals (wild type rise time, 1.5±0.2 milliseconds; n=9; decay time, 17.4±3.0 milliseconds; n=7) (Tg-2 rise time, 1.0±0.1 milliseconds; n=7; P<0.05; decay time, 9.6±0.9 milliseconds; n=7; P<0.05). Moreover, the synaptic delay was significantly shorter in Tg-2 (1.4±0.1 milliseconds; n=7) compared with wild-type synapses (1.9±0.2 milliseconds; n=9; P<0.05). The large and consistent differences in synaptic delay (about 500 μseconds) observed between wild-type and Tg-2 mice are unlikely to be attributable to changes in the distance between stimulating and recording electrodes, since the stimulating electrode was always placed within 100±20 μm from recorded DCN cells and, with an estimated conduction velocity of PC axons of 0.7 m/second (Clark et al., 2005), the jitter of delay would be only ±28 μseconds. Although the possibility of different conduction velocity or axon tortuosity cannot be excluded, no changes in synaptic morphology supporting a longer or slower diffusion of GABA were detected in transgenic mice (see Fig. 2C and Table 1). These data imply that the time needed to release a SV is reduced, consistent with an involvement of pepE in the reactions that determine kinetics and synchronization of quantal release.

To ascertain whether the changes in synaptic physiology observed in PC-DCN synapses were associated with differences in synaptic structure and/or SV distribution, PC terminals making axo-somatic synapses on nuclear neurons were analyzed by electron microscopy. PC terminals from transgenic mice showed a marked decrease in the total number of SVs compared with wild-type terminals (Fig. 9A and Table 1). In the absence of detectable differences in both the average synaptic area and active zone number and length, the total number of SVs per synapse in Tg-2 mice was ∼55% of that of wild-type mice.

The extent of clustering of SVs in PC nerve terminals from transgenic and control mice was deduced from the mean intervesicular distance (MID) corrected for the SV density (clustering index, see Materials and Methods). Despite a decrease in the number of SVs, the overall MID was lower in transgenic mice, resulting in a significantly higher clustering index with respect to wild-type mice (Table 1). In addition, SVs were more concentrated in proximal shells, as shown by a significant decrease in the mean distance from the active zone (Table 1 and inset of Fig. 9B).

A detailed morphometric analysis of SV distribution within PC terminals (Fig. 9B) indicated that the decrease in SV density was highly significant at distances between 400 and 1400 nm from the active zone. These data are consistent with previous results obtained after injection of pepE in invertebrate excitatory synapses and can explain the decreased response in GABA release to ionomycin observed in transgenic animals. Interestingly, SVs within 200 nm of the active zone were totally preserved or even slightly increased, reaching significance (P<0.05) in the 50-100 nm shell.

To investigate the functional role of the highly conserved domain E of synapsin in a central inhibitory synapse, we selectively overexpressed the most conserved sequence of the domain in cerebellar PC in two independent transgenic lines. The recombinant peptide was expressed at detectable levels, was virtually absent from PC somata and was correctly targeted to nerve terminals in the presence of endogenous synapsins, consistent with a positive role of this domain in synapsin targeting to nerve terminals (Gitler et al., 2004a). Despite an impaired motor performance, transgenic mice had a normal cerebellar structure, PC morphology and density of PC terminals onto DCN neurons. Interestingly, PC terminals of L7-pepE mice displayed an accelerated kinetics of exocytosis and a marked resistance to depression evoked by high-frequency stimulation. This increased synaptic performance was associated with a marked decrease in the number of SVs away from the active zone, whereas SVs close to the active zone were preserved or slightly increased.

The phenotype of L7-pepE mice points to a specific role of the synapsin domain E in the regulation of SV trafficking at both pre- and post-docking steps, involving both the RP and the RRP of SVs. In both the squid giant synapse (Hilfiker et al., 1998) and mammalian PC-DCN synapses (this study), overexpression of pepE induced a marked depletion of SVs in the RP, leaving the RRP virtually unaffected. This effect is in agreement with the inhibition of the actin and synapsin interactions by pepE recently reported by our laboratory (Hilfiker et al., 2005), which might induce induce loss of SVs bound to the actin cytoskeleton.

On the other hand, the acceleration of SV cycling and the kinetics of SV fusion found in the PC-DCN synapses is opposite to that observed in the squid giant synapse. Although these observations strengthen a central role of domain E in the postdocking steps of neurotransmitter release, they reveal a more complex role of domain E in mammalian central inhibitory synapses. Thus, it is possible that pepE accelerates SV cycling within the RRP by interfering with the interactions of SV-associated synapsins with the actin-rich cytomatrix of the active and periactive zones (Bloom et al., 2003; Hilfiker et al., 2005), which, in mammalian synapses, might slow down SV cycling. This interpretation is supported by the observation that, in hippocampal nerve terminals, the disruption of the actin cytoskeleton resulted in both synapsin dispersion and faster SV cycling (Sankaranarayanan et al., 2003). However, the possibility that that peptide E interferes with a synapsin interactor that negatively regulates the rate of release cannot be excluded. Interestingly, the C-terminal region of synapsin has been reported to interact with Rab3A, a SV-associated GTPase regulating the final steps of release (Geppert et al., 1997; Giovedì et al., 2004a; Giovedì et al., 2004b; Schluter et al., 2006).

The phenotype of L7-pepE mice also confirms that the relative functional importance of the RP and RRP of SVs in sustaining release varies between neurons with different patterns of electrical activity. During short high-frequency trains that are likely to involve `kiss-and-run' mechanisms and fast SV recycling pathways (Valtorta et al., 2001), mutant mice showed both a lower depression rate and a higher steady-state response than wild-type mice. Although the former effect might reflect lower release probability, this would imply no change in the steady-state level (Brenowitz and Trussell, 2001). The concomitant higher steady state observed suggests that the lower depression is due to an acceleration of SV cycling that allows faster refilling and slower depletion of the RRP.

It has been recently reported that mice lacking all synapsin isoforms have distinct deficits in excitatory and inhibitory synapses. Thus, glutamatergic terminals display normal release evoked by single stimuli and enhanced depression, whereas GABAergic terminals exhibit a decreased release in response to single stimuli and no effect on depression (Gitler et al., 2004b). Indeed, the recycling capacity of the RRP tends to be adapted to the activity pattern and is higher in terminals undergoing tonic high-frequency activity than in terminals experiencing infrequent phasic activation. Although the latter terminals take advantage of a large RP during bursting activity, the former have to rely on efficient recycling mechanisms (for reviews, see Brodin et al., 1997; Rizzoli and Betz, 2005). In vivo, PC neurons typically discharge at high frequencies (Thach, 1968; Thach, 1970; McDevitt et al., 1987). Thus, because of the very high basal firing rate, PCs are likely to rely predominantly on efficient recycling of the RRP rather than on SV recruitment from the RP. This is also suggested by the similar lower depression found in mutant mice during both short and sustained high-frequency trains, in the presence of a substantial depletion of the RP. Such RP depletion becomes apparent only under conditions of massive exocytosis that release SV from all nerve-terminal pools.

The data presented here demonstrate that overexpression of domain E increases synaptic strength at PC-DCN inhibitory synapses and participates in the motor behaviour phenotype observed in transgenic mice. These results might have important implications as to the role of synapsins in the regulation of central inhibitory synapses and network excitability. Synapsin knockout mice display a strong epileptic phenotype attributable to a specific deficit in inhibitory transmission (Terada et al., 1999; Gitler et al., 2004b). Moreover, the recently reported X-linked human epilepsy associated with a non-sense mutation of the SYN1 gene involves a truncation of the synapsin I C-terminal region that includes domain E (Garcia et al., 2004). It is tempting to speculate that the absence of this domain plays a crucial role in the impairment of synaptic inhibition leading to epileptiform activity.

A minigene (PepE) was made by hybridization of two complementary oligonucleotides producing a double stranded product with 5′ overhangs compatible with BamHI restriction sites and coding for the last 25 residues of the rat synapsin Ia domain E. The sequences of the two oligonucleotides are as follows: PepE sense, 5′-GATCCGCCACCATGAGCCTGAGCCAGGACGAGGTGAAGGCCGAGACCATCAGAAGCCTGAGAAAGAGCTTCGCCAGCCTGTTCAGCGACTAAG-3′; PepE antisense, 5′-ATCCTTAGTCGCTGAACAGGCTGGCGAAGCTCTTTCTCAGGCTTCTGATGGTCTCGGCCTTCACCTCGTCCTGGCTCAGGCTCATGGTGGCG-3′. The L7-pepE vector was constructed by inserting this fragment into the BamHI site of pL7DAUG in which all the potential start codons were eliminated (De Zeeuw et al., 1998; Zhang et al., 2002). Thus, translation could only be initiated from the start codon provided within the inserted minigene, which was optimized with a Kozak sequence for efficient translation in mammalian cells (Kozak, 1986). The L7-PepE transgene, digested with HindIII/EcoRI, was injected into fertilized mouse eggs. Four positive transgenic lines were identified by PCR analysis and two of them (Tg-1 and Tg-2) were crossed to homozygosity. All animal procedures were approved by the Animal Care Committee of the University of Genova and by the Italian Ministry of Health.

RNA was extracted from brain tissue using TRI Reagent (Sigma). Reverse transcription was conducted with random examers and SuperScript II Reverse transcriptase (Invitrogen) as described previously (Corradi et al., 2003). Primers (forward, 5′-CAGGCCAGAACCCAGAAAG-3′; reverse, 5′-ATCCTTAGTCGCTGAACAGG-3′) specific for the cDNA region encoding the L7-PepE were used for PCR amplification. Hypoxanthine guanine phosphoribosyltransferase (HPRT) primers used for normalization were 5′-TCCCTGGTTAAGCAGTACAG-3′ (forward) and 5′-GACGCAGCAACTGACATTTC-3′ (reverse).

Antisense oligonucleotides of 45 or 30 bases in length specific for the cDNA region encoding the L7-PepE were labelled at the 3′ end using [³⁵S]ATP (Amersham) and terminal deoxynucleotidyl transferase (Roche). The hybridization of brain cryostat sections (14 μm thick) was carried out as previously described (Zoli et al., 1995). Briefly, sections were fixed with 4% paraformaldehyde (PFA), washed in 0.1M phosphate-buffered saline (PBS), acetylated and delipidated in ethanol and chloroform. After a 3 hour prehybridization in 0.6 M NaCl, 0.01 M dithiothreitol, 0.1 M Tris-HCl (pH 7.4), 0.05 M EDTA, 5% Denhardt's solution, 1 mg/ml Poly(A) (Roche), 10 mg/ml t-RNA, 1 mg/ml herring sperm DNA (Promega), 50% formamide at 37°C, sections were hybridized for 20 hours at 37°C with the labelled oligonucleotides. Sections were washed twice in 2× standard citrate solution (SSC, 3 M NaCl, 0.3 M sodium citrate) at room temperature, four times in 2× SSC at 54 or 48°C and twice in 1× SSC at room temperature. Sodium thiosulfate (63 mM) and β-mercaptoethanol (14.3 mM) were added to all washing solutions. After rinsing in ice-cold distilled water and drying, sections were first exposed to low-resolution autoradiographic film (Amersham) and then to high-resolution photographic emulsion (Kodak) for 1-2 months.

Mice were perfused under deep anesthesia (33 mg/kg xylazine and 33 mg/kg ketamine) with 0.1 M PBS followed by 4% paraformaldehyde (PFA) in 0.1 M PBS. After perfusion, brains were postfixed for 2 hours in 4% (w/v) PFA and transferred to 15% (w/v) sucrose in 0.1 M PBS and incubated overnight at 4°C. Cryostat sections (14 μm thick) were mounted on superfrost slides (Menzel-Glaser) and stored at –80°C for up to 15 days. Immunohistochemistry was performed as previously described (Fassio et al., 2000) using the following antibodies: monoclonal anti-calbindin D-28K (Sigma), anti-VAMP2 (Synaptic Systems), polyclonal anti-synaptophysin (G111), anti-synapsin I (G177) and anti-synapsin pepE (G304) raised in our laboratory (Valtorta et al., 1988; Pieribone et al., 1995; Vaccaro et al., 1997). Primary antibodies were detected using either the avidin-biotin method with diaminobenzidine as a chromogen (Vector Laboratories) or Alexa Fluor 488- and 546-conjugated secondary antibodies (Invitrogen). Specimens were viewed with an epifluorescence Olympus inverted microscope. Images were recorded with a Hamamatsu C4742-98 ORCA II camera and processed using Image Pro Plus 4.5 (Media Cybernetics). DCN sections, double-labeled with anti-calbindin and anti-synaptophysin antibodies, were subjected to confocal scanning microscopy (Biorad MRC1024). Stacks of images were acquired in the DCN region at 1.5 μm intervals along the z-axis. Based on synaptophysin staining, synaptic masks were generated by using the granulometric algorithm (Prodanova et al., 2006). The synaptic signal of calbindin was obtained by applying the synaptic mask to the calbindin signal. Nerve terminal counts were performed by using the `analyze particle' function of the Image J program (NIH, Bethesda, MD).

Primary cultures of PC were prepared from E18 Tg-2 mice following the reported procedure (Tabata et al., 2000). Cultures were either subjected to immunocytochemistry with anti-calbindin and anti-pepE antibodies or harvested, solubilized in stop solution and subjected to SDS-PAGE in Tris-Tricine (Schagger and von Jagow, 1987) and immunoblotting of PVDF membranes with anti-pepE antibodies.

Primary hippocampal neurons obtained from E18 wild-type embryos (Banker and Cowan, 1977) were transfected at 5 days in vitro (DIV) with either pEGFP, pEGFP-synapsin Ia or pEGFP-domain E vector using Effectene (Qiagen). Expression and targeting of GFP or of the GFP chimeras were followed at 7-9 DIV by fluorescence microscopy of neurons labelled with anti-VAMP2 antibodies.

Mice (3 months old, weighing 20-25 g) housed under standard temperature, humidity and light/dark cycle underwent the following tests during the light phase:

The mouse was placed on a round (2 cm diameter, 50 cm length) rod 35 cm elevated from the bench. The latency to fall was measured in three separate trials (with a 5 minute intertrial interval). Each trial ended at 180 seconds.

The mouse was placed facing up on a pole (2 cm diameter, 50 cm long). The end of the pole was then gradually lifted to a vertical position and the time the mouse stayed on the pole was recorded for a maximum of 60 seconds. Time values were converted to a pole test score as follows: stayed on 60 seconds and climbed halfway down the pole=8; climbed to the lower half of the pole=9; climbed down and got off the pole in 51-60 seconds=10, 41-50 seconds=11, 31-40 seconds=12, 21-30 seconds=13, 11-20 seconds=14, 1-10 seconds=15 (McIlwain et al., 2001). Scores lower than 8 refer to animals that fall from the pole, an event that never occurred with transgenic animals. Each animal was tested in three consecutive trials.

Mice were tested with the rota-rod (Ugo Basile, Via G. Borghi 43, 20125 Comerio, VA, Italy) under conditions of either constant speed or constant acceleration (Nolan et al., 2003). In the constant-speed test, mice were first trained until they could remain on the rod at 4 rpm for three consecutive 120-second trials. The next day the mice were placed back on the rod for a trial at 40 rpm. The time a mouse could remain on the rota rod (maximum 120 seconds) was recorded and the process was repeated for speeds of 30, 20 and 4 rpm. Animals were given four trials per day with a 1-hour inter-trial interval. In the accelerating test, mice were placed on the rota rod starting at 4 rpm, slowly accelerating to 40 rpm. The maximum observation time was 5 minutes. Animals were tested for 3 consecutive days, receiving four trials on day 1 and three trials on both days 2 and 3, with a 1-hour inter-trial interval.

Mice were put in a dark tunnel (7 cm wide, 50 cm long, 10 cm high) with white paper on the floor. Before traversing the tunnel, the hindpaws of the animals were dipped in non-toxic black ink (Feil et al., 2003). Footprints were analyzed using the `Footprint 1.22' software (Klapdor et al., 1997) to calculate stride width, stride length and maximum difference in stride-length values.

Mice were killed by decapitation and DCN were rapidly dissected out. Percoll-purified synaptosomes were prepared as previously described (Dunkley et al., 1988). Synaptosomes were resuspended in physiological medium with the following composition (mM): NaCl 125, KCl 3, MgSO₄ 1.2, CaCl₂ 1.2, NaH₂PO₄ 1, NaHCO₃ 22, glucose 10 (pH 7.4 when equilibrated with 95% O₂ and 5% CO₂); and incubated at 37°C for 15 minutes in the presence of 0.04 μM [³H]GABA. After labelling, aliquots of the suspensions (<100 μg protein/filter) were stratified onto microporous filters at the bottom of parallel superfusion chambers maintained at 37°C (Raiteri and Raiteri, 2000) and superfused with standard medium at 0.5 ml/minute. Under these conditions, synaptosomes constitute less than a monolayer and indirect effects mediated by compounds released by neighbouring particles are virtually absent. After a 33-minute equilibration period, four 3-minute fractions were collected. Synaptosomes were exposed to a 90-second pulse of high KCl (12, 15 or 35 mM) or ionomycin (0.5 mM) at the end of the first fraction collected (t=39 minutes). In some experiments, three successive high-KCl pulses were applied to the same synaptosomal preparation at 18-minute intervals (t=39, 57 and 75 minutes). Aminooxyacetic acid (50 μM) was present throughout the experiment to prevent [³H]-GABA metabolism. Collected samples and filters were counted for radioactivity. The amount of radioactivity present in each sample was calculated as fractional rate and the stimulus-evoked overflow was estimated by subtracting basal release from the radioactivity measured in the samples collected during and after the stimulation pulse.

Whole-cell patch-clamp recordings from cerebellar slices were performed following standard procedures (D'Angelo et al., 1995; Hansel et al., 2001; Sola et al., 2004). Parasagittal slices (220 μm thick) were taken from the cerebellar vermis of P8-P12 mice decapitated after deep halothane anesthesia. During the slicing procedure, the cerebellar vermis was immersed in a cold solution containing (mM): K-gluconate 130, KCl 15, EGTA 0.2, HEPES 20, Glucose 10 (pH 7.4 with NaOH). Before recording, slices were incubated at 32°C for at least 30 minutes in oxygenated bicarbonate-buffered saline (standard Krebs solution) containing (mM): NaCl 120, KCl 2, MgSO₄ 1.2, NaHCO₃ 26, KH₂PO₄ 1.2, CaCl₂ 2, glucose 11 (pH 7.4 when equilibrated with 95% O₂ and 5% CO₂). Slices were then transferred into the recording chamber and perfused at 1.5 ml/minute with oxygenated Krebs solution at room temperature (20-23°C). The Krebs solution was added with 50 mM D-APV (D-2-amino-5-phosphonovaleric acid) and 20 mM CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) (Tocris). Slices were visualized with an upright epifluorescence microscope (Zeiss Axioskop 2FS), equipped with a ×63 water-immersion objective (0.9 NA) and DIC optics, using infrared illumination (illumination filter 750 nm) and an IR CCD camera (T.I.L.L. Photonics). Recordings were obtained using a MultiClamp 700A amplifier (Molecular Devices).

Whole-cell recordings were performed from large spheroidal neurons (maximum diameter 15-30 μm) identified with DIC-IR Nomarsky interference contrast. These neurons showed spontaneous firing at >10 Hz in cell-attached before obtaining the whole-cell configuration. The input resistance measured by recording the response to –10 mV voltage steps delivered from the holding potential of –65 mV was 440.9 54.1 MΩ (n=10). Thus, these neurons corresponded to type-I DCN projection neurons according to the previously reported classification (Czubayko et al., 2001) (see also Anchisi et al., 2001). The patch-pipette contained (mM): Cs₂SO₄ 81, NaCl 4, MgSO₄ 2, CaCl₂ 0.02, BAPTA 0.1, glucose 15, ATP-Mg 3, GTP 0.1, HEPES 15, QX314 1.5. This solution maintained resting free [Ca²⁺] at 100 nM and pH was adjusted to 7.2 with CsOH. Patch-clamp pipettes filled with this solution had a resistance of 1-2 MΩ before seal formation. After obtaining the whole-cell configuration, membrane potential of DCN neurons was kept at –65 mV. PC axons were stimulated via a stimulus isolation unit with a bipolar silver wire, whose active filament was inserted into a glass pipette (10 μm tip diameter) filled with standard Krebs solution. The stimulating electrode was placed at 100 20 μm from the recorded DCN neuron. IPSCs were elicited at a basal frequency of 0.1 Hz with constant stimulation intensity. Two types of high-frequency trains were applied: (1) 200 milliseconds at 100 Hz to analyze short-term plasticity regulated by RRP and (2) 120 seconds at 50 Hz to investigate the involvement of the RP. IPSC peak value was calculated as the difference between peak and base (see Fig. 5). Under the same conditions, PC spike firing was monitored using loose cell-attached recordings that do not perturb basal discharge frequency (Forti et al., 2006).

Chopper sections containing DCN from three wild-type and three Tg-2 mice were fixed for 30 minutes at room temperature with 2% PFA, 1% glutaraldehyde, 1% sucrose in cacodylate buffer (100 mM, pH 7.4), washed five times in cacodylate buffer, and postfixed for 1 hour with 1% OsO₄ in the same buffer. Samples were subsequently washed five times for 5 minutes with cacodylate buffer, dehydrated and embedded in Epon 812 (Ceccarelli et al., 1973). Silver-gray sections were cut on a microtome (Reichert-Jung Ultracut), stained with 4% uranyl acetate, 0.4% lead citrate, and examined in a Hitachi H-7000 electron microscope. The identification of the PC-DCN synapses was carried out based on the following criteria: (1) Gray type II `symmetrical' synapses; (2) elliptical and dispersed SVs; (3) location on soma or proximal dendrites of DCN neurons (Chan-Palay, 1977). All the encountered synapses whose features made them recognizable as PC nerve endings were analyzed. The geometrical parameters of the nerve endings as well as the SV density measured in samples from wild-type mice were virtually identical to those reported (Chan-Palay, 1977), indicating that the sample was representative of the whole population of PC nerve endings.

Digitized electron micrographs of PC-DCN synapses were subjected to computer-assisted morphometric analysis to calculate the following parameters: nerve terminal area, length of the synaptic contact, number and length of active zone, number of SVs and, for each SV, the distance from the closest active zone and the mean distance from all surrounding SVs (mean intervesicular distance, MID, a measure of SV clustering). Since MID depends on SV density, a clustering index was obtained, corresponding to the ratio between the MID calculated for a simulated homogeneous SV distribution and the measured MID. The area occupied by mitochondria (which was similar in the two experimental groups) was subtracted from the total area. Parameters were calculated from the EM images of 60 synapses from three wild-type mice and 54 synapses from three transgenic mice. Within each animal, individual PC-DCN synapses were analyzed to give mean individual values that were subsequently averaged within the same experimental group. To analyze SV distribution with respect to the active zone, the calculated distances of SVs from active zone within each nerve terminal were grouped into 50 classes of 50 nm. From the frequency of SVs whose distance from the active zone was falling in the various classes, frequency distribution histograms were obtained for each nerve terminal and were subsequently averaged within the same experimental group.

Data are reported as mean ± s.e.m. and statistical analysis was performed using the program Prism (GraphPad Software). Data were analyzed using one-way ANOVA followed by either the Student's t-test or multiple comparison tests (either Dunnett, Duncan or Newman-Kleus test). Frequency distribution histograms were analyzed by the Kolmogorov-Smirnov normality test followed by either parametric or non-parametric tests (Student's t-test and Mann-Whitney U test, respectively). Both types of analysis gave fully comparable significance levels.

We thank P. Greengard (The Rockefeller University, New York, NY) for useful discussion and critical reading of the manuscript, L. Pozzi and C. Tiveron (Telethon Transgenic Mice Service Center, Istituto Regina Elena, Roma, Italy) for embryo injections, M. Zoli and A. Zanardi (University of Modena, Italy) for help with in situ hybridization experiments, D. Ghezzi (Politecnico di Milano, Italy) for help with the morphometric analysis. This work was supported by grants from the Italian Ministry of University (Cofin 2004 and 2005, FIRB `Neuroscienze' and `Postgenomica' to F.B., G.B., E.D'A. and F.V.), Fisher Foundation for Alzheimer's Disease Research, Associazione Italiana Ricerca sul Cancro, Fondazione Franco e Luisa Mariani and Consorzio Italiano Biotecnologie (to F.B.), Consiglio Nazionale delle Ricerche (Progetti Strategici Neuroscienze and Genomica Funzionale to F.B. and F.V.) and by projects of the European Community (Cerebellum QLG3-CT-2001-02256 and Spikeforce IST-2001035271 to E.D'A). The financial support of Telethon-Italy (grant no. GGP05134 to F.B and F.V.) is gratefully acknowledged.